Chapter IV: Abrupt transitions and hysteresis in glacial ocean circulation
4.1 Introduction
During the last glacial period there were a series of abrupt changes in Northern Hemisphere high-latitude temperature, as recorded in Greenland ice cores (Dans- gaard et al., 1993; NGRIP members, 2004). Ice core records from Greenland and Antarctica show that changes in inferred temperature during the last glacial period were not in phase between the hemispheres (Blunier and Brook, 2001). During cold periods in Greenland (known as “stadials”), temperature in Antarctica gradu- ally rose, and during warm periods in Greenland (“interstadials”), temperature in Antarctica gradually fell (Figure 4.1) (Blunier and Brook, 2001). This observation has been coined the bipolar seesaw (Broecker, 1998; Thomas F Stocker and Johnsen, 2003; T F Stocker, 1998).
The simplest mechanism to explain this phenomenon is the asymmetry of equator- to-pole heat transport by the ocean in the Atlantic (Crowley, 1992). Because deep water is formed in the North Atlantic, there is net transport of warm surface water north across the equator and net transport of cold deep water south. In order to match the abrupt transitions in the north with the gradual transitions in the south, a heat reservoir was added to the simple bipolar seesaw model to damp the signal in the south (Thomas F Stocker and Johnsen, 2003). This “thermal” bipolar seesaw was able to qualitatively match ice core records from Greenland and Antarctica. Recent efforts to more precisely align high-resolution Northern and Southern Hemisphere ice core records have allowed for a better determination of the interhemispheric phasing at these abrupt climate change events (WAIS Divide Project Members, 2015). WAIS Divide Project Members (2015) found a Northern Hemisphere lead of
∼ 200±100 years at these abrupt transitions. However, the origin of this 200-year timescale has been somewhat difficult to explain because it is slow for a typical atmospheric adjustment timescale, O(≤1 yr), but short for a full diffusive ocean adjustment timescale,O(1000 yr) (Zhengyu Liu and Alexander, 2007).
The global overturning circulation involves a central role for the Southern Ocean
87
0 20 40 60 80 100 120
Age (kyr BP)
MIS 1 MIS 2 MIS 3 MIS 4 MIS 5
-54 -52 -50 -48-46 -44 -42 -40
δ18O EDML (‰) -46 -44 -42 -40 -38 -36 -34 -32
δ18O NGRIP (‰)
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Pa/Th
2
34 56 7 8 9
10 11 12 13
1415 1617 18
19 20 21
A2 A3 A4 A5 A6
A7 A1
A
B C
Figure 4.1: Greenland and Antarctic δ18O records with Pa/Th circulation tracer.
Oxygen isotopic composition of NGRIP ice core in Greenland (A) and EDML ice core in Antarctica (C). The δ18O composition of ice is a proxy for regional temperature. Numbers mark the Dansgaard-Oeschger (DO) events in Greenland and corresponding AIM events in Antarctica. B) Compilation of Pa/Th records from the Bermuda Rise (youngest section from McManus et al. (2004) (diamonds), middle section from Lippold et al. (2009) (circles), oldest section from Böhm et al.
(2015) (squares)). Pa/Th ratio is a proxy for the strength of North Atlantic Deep Water (NADW). Horizontal dashed line marks the Pa/Th production ratio of 0.093.
Bar at top shows the Marine Isotope Stage boundaries.
(J. Marshall and Speer, 2012; Kuhlbrodt et al., 2007). In the Southern Ocean, strong westerly winds drive the vigorous Antarctic Circumpolar Current (ACC), the largest current on the planet. However, unlike the rest of the ocean, these strong zonal winds cannot drive mean geostrophic meridional flow because there are no continents to create sustained pressure gradients. Therefore, other transport mechanisms are required. Over the past decade, great strides have been made in our understanding of Southern Ocean dynamics and circulation. Notably, we now have a theoretical framework that describes circulation as a balance between Southern Ocean westerly winds, that steepen isopycnals via Ekman processes, and eddies, that act to relax them (J. Marshall and Radko, 2003).
The wind-driven, or mean, component of the overturning circulation is defined as ψ¯ = −τ/ρ0f, where τ is the surface wind stress, ρ0 is the mean ocean density, and f is the Coriolis parameter. The counteracting eddy-driven component can be parameterized as ψ∗ = K s, where K is the eddy diffusivity and s is the isopycnal
slope. At steady-state, the balance between wind and eddies leads to a structure where isopycnals slope from the surface to depth across the Southern Ocean, and the surface meridional distribution of density classes is mapped into a vertical stratification at the northern edge of the ACC. Importantly, though, this steady-state isopycnal structure does not necessarily allow for meridional overturning—in fact if ¯ψ perfectly matches ψ∗, then there is no net meridional transport. The balance between the wind and eddy-driven circulation is known as the residual circulation:
ψr es =ψ¯ +ψ∗. Residual flow occurs largely along isopycnals in the subsurface and then crosses isopycnals in the mixed layer as a result of surface buoyancy fluxes (J. Marshall and Radko, 2003; Speer, Rintoul, and Sloyan, 2000). Where there is a positive surface buoyancy flux, water in the mixed layer can be converted from a heavier density class to a lighter density class, and vice versa. Because the residual circulation above the sill depth occurs along isopycnals in the ACC, it allows for adiabtic pole-to-pole upwelling along the subset of isopycnals that outcrop both in the Southern Ocean and in the North Atlantic (Wolfe and Cessi, 2011). In a zonally averaged depiction of the meridional overturning circulation, this adiabatic circulation is the upper circulation cell. The lower cell is necessarily diabatic because it involves isopycnals that only outcrop in the Southern Ocean.
In the modern ocean, we observe that closure of the overturning circulation requires water mass modification in both the Atlantic and Indo-Pacific basins (Lumpkin and Speer, 2007; Talley, 2013). This is because North Atlantic Deep Water (NADW) is dense enough that it upwells in a region of the Southern Ocean where it experiences a negative surface buoyancy forcing. This water therefore moves to the south and sinks again as Antarctic Bottom Water (AABW), part of the lower circulation cell.
If this AABW flows back into the Atlantic basin, then it upwells back into NADW.
It is only once AABW flows into the Indian and Pacific basins that it is able to diffusively upwell to a density class where it can close the overturning loop. It does this either via Pacific Deep Water (PDW), which upwells in a positive buoyancy forcing region of the Southern Ocean and returns north as Antarctic Intermediate Water (AAIW), or via the Indonesian Throughflow, flowing first from the surface of the Western Equatorial Pacific into the Indian Ocean, and then from the Indian Ocean into the Atlantic through the Agulhas Leakage (Talley, 2013).
Building upon the work of Curry and Oppo (2005), Talley (2013), and J. Marshall and Radko (2003), Ferrari et al. (2014) recently put forth a theoretical model for changes in overturning circulation structure at the Last Glacial Maximum (LGM)
89 compared to today. This model relies on two key observations: (1) that the change from positive surface buoyancy forcing to negative surface buoyancy forcing aligns roughly with the quasi-permanent sea ice edge (defined as ice-covered 70% of the time) (Ferrari et al., 2014), and (2) that there is dramatically higher diapycnal mixing over rough bottom topography (Polzin et al., 1997). The Ferrari et al.
(2014) model also relies on the simplification of a constant slope for isopycnals in the Southern Ocean (approximately equal to τ/ρ0f K). Using the observation that the boundary between the positive and negative buoyancy forcing regions (and thus the upper and lower circulation cells) lies at the summertime sea ice edge and the approximate isopycnal slope, Ferrari et al. (2014) calculate the depth of the boundary between the upper and lower circulation cells in the northern basins. For the modern ocean this depth comes out to∼2200 m, below the mean depth of the mid-ocean ridges (2000 m). If sea ice was expanded at the LGM, as is implied by data and models, then this geometric argument predicts a shoaling of the boundary between the upper and lower circulation cells. Ferrari et al. (2014) estimate a 5◦ latitude shift of the summertime sea ice edge (based on output from the National Center for Atmospheric Research Community Climate System Model version 3), which would lead to a 500-m shoaling of the boundary between the upper and lower circulation cells—enough to raise it above the region of rough topography in the ocean and reduce diapycnal mixing. One potential result of this change in diapycnal diffusivity, driven by a shoaling of the boundary between the circulation cells, is a change from an intertwined “figure-eight” overturning circulation, as described by Talley (2013), into a two-cell overturning circulation, where there is less connection between upper and lower circulation cells and more potential to sequester CO2 in the deep ocean.
In order to understand the complex interplay of physics and biogeochemistry that lead to glacial-interglacial cycles, we rely on models as well as paleoceanographic reconstructions, but it can be difficult to accurately model processes that occur on millennial timescales. On one end of the modeling continuum are general circula- tion models (GCMs), which attempt to capture all the physics of the ocean (and/or atmosphere), but are faced with the dilemma of run duration versus model resolu- tion. It is widely understood that small-scale processes are crucial for accurately representing the Southern Ocean residual circulation and Antarctic Bottom Water formation, yet the computational cost is prohibitive to run these high-resolution models that explicitly resolve these processes for thousands of years. The most recent and comprehensive GCM modeling effort to understand ocean dynamics
during the most recent deglaciation, the TraCE simulation (Z Liu et al., 2009), ran the Community Climate System Model, version 3 (CCSM3) for 11,000 years with a nominal ocean resolution of 3◦(for comparison, 1/4–1/6◦models are considered
“eddy permitting” and 1/10◦are considered “eddy resolving”). On the other end are simple theoretical models, that parameterize fluxes based on theoretical relations rather than attempting to resolve them based on the equations of motion. An early and well-known example of this type of model is from Gnanadesikan (1999), who sought to understand the depth of the main pycnocline. His model was essentially 1-dimensional, with two layers and a single basin. Subsequent efforts by D. P. Mar- shall and Zanna (2014) expanded the Gnanadesikan model to multiple layers, but kept the same general model geometry and did not attempt to include Antarctic Bot- tom Water formation. Another example of a simple physical model is the dynamical box model of Goodwin (2012), which looks to solve the problem of prescribed box volumes and fluxes in biogeochemical models.
Recently, a simple but powerful new model has been developed by Thompson, Andrew L. Stewart, and Bischoff (2016). It uses residual mean theory in the Southern Ocean (J. Marshall and Radko, 2003), but keeps the Atlantic and Pacific basins separate, unlike most models, which reduce dimensionality by zonally averaging.
This added dimension enables the model to occupy the figure-eight circulation configuration of Talley (2013). It is an isopycnal model with very coarse resolution—
density surfaces are flat-lying in the Atlantic and Pacific basins, and slope to the surface (with a constant slope) in the ACC. Since there are two basins (the Atlantic and Indo-Pacific), the model allows exchange via zonal convergence or divergence in the ACC and through a simulated Indonesian Throughflow. Each density surface can be defined by two points: a depth, z, in the basin, and an outcrop position, y, in the ACC. Thompson, Andrew L. Stewart, and Bischoff (2016) choose four density classes (three isopycnal surfaces), because that is the minimum number that allows two overturning cells to close within a single basin. This means that the model can occupy either the figure-eight configuration or the two-cell configuration.
They solve both the overturning circulation and the stratification at steady-state by solving for each isopycnal depth and outcrop position in each basin and the zonal convergence. The model also calculates the water mass modification and transport from the basin stratification (z) and ACC outcrop position (y).
In this chapter we modify the Thompson, Andrew L. Stewart, and Bischoff (2016) model in order to investigate the ocean response to transient changes in circulation.
91 In particular, we simulate the abrupt changes in NADW formation rate associated with the Heinrich and Dansgaard-Oeschger events during the last glacial period (Figure 4.1) (Dansgaard et al., 1993; Böhm et al., 2015). In these simulations, we look at the timescale associated with communicating the signal of an abrupt change in NADW strength from north to south. The transient model response to abrupt NADW flux changes occurs much faster than a full diffusive ocean adjustment, and provides a mechanistic explanation to the observations of WAIS Divide Project Members (2015) (Thompson, Hines, and Adkins, submitted). Then we perform a series of experiments to determine which regions of parameter space allow for abrupt transitions in basin stratification and circulation configuration, focusing on how the density of NADW and the strength of NADW interact. We compare our results to the predictions made by Ferrari et al. (2014).